Forum index > Bug report ✗

Freq ratio key depending, normal?

Merkur
When modulating a carrier, I noticed that the harmonics are slightly moving on some notes when the sound is perfectly steady on others. Actually the frequency ratio between modulator and carrier is slightly key dependent: using an frequency analyzer you can really see it.
Is it intended? It makes the sound a bit alive, but sometimes you don't want that effect...
KODAMO
Can you attach a recording of it ? I'm not aware of this behavior, although it could be a characteristic of our FM engine
Merkur
Ok I'll do that tomorrow!
Merkur
First sorry for my sloppy formulation : harmonics are note "moving" but are "beating". (That normally should only happen when negative frequency side-bands correspond with positive freq. s.b., and with ratio "slightly" irrational, thus moving the "fine" frequency parameter.)

It happens with ratios C/M 1/2 and 2/1 (very noticeable beating playing C5, C5#, D5, G5, A5, Bb5, B5 but no problem with Eb5, E5, F5, F5#, G#5. (The behavior changes in other octaves...).
No problem with 1/1...
That is strange because those 3 C/M ratios are sharing the characteristic of having negative and positive side-bands sharing the same "harmonics", so why no problem with 1/1?

I am joining an audio files playing a patch with no velocity, square envelopes for C (operator 1) & M (op2), C/M = 1/2 and Modulation Index of 50. I played only one octave as an exemple (C5 to C6), but the behavior is similar in other octaves.
KODAMO
Thank you for the recording.
This is happening because of the frequency resolution the EssenceFM uses. Frequencies are stored as large numbers, no matter how large they are they would need to be of infinite precision to allow for perfect C/M relationships.

Depending on the note you're playing, you'll get perfect or imperfect C/M ratios.

Example with 2:1 ratio (with arbitrary numbers)

OP1 freq: 100000000
OP2 freq: 50000000
=> No beating

OP1 freq: 100000001
OP2 freq: 50000000
=> Beating


Would be cool to see how Yamaha FM chips (or other gear) behave.
Not sure if that's an issue, perfect digital representations rarely lead to better sound. But that's intersting to know what happens inside!
Merkur
Thank you for your quick reaction! I understand.
I will check my TG77 to compare. Actually it is so boring to program that beast that I never really experimented FM as deeply as I am doing for two weeks now with your wonderful synth!
I'll keep you informed.
Merkur
I just checked on the TG77 : it is indeed absolutely stable...
On the EFM I noticed that it is possible to compensate manually by moving the "fine" parameter periodically to "-1" in order to get back in tune... Maybe a clue to imagine an tuning interpolation system?
Going back to what you said about the EFM : as you are using memory to store the frequencies you will always have approximations compared to actually compute the operator frequency from a "master" operator. But I suppose that would mean completely rethink your algorithm...
Anyway that's the proof that even a digital instrument based on a numerical system can have his own specific character!
(And if I really want to avoid that artifact I will compose music only with Eb, E, F, F# & G : limitation stimulates creativity ! ;-) Just kidding! It is really a wonderful instrument that MUST be better known!
I will definitely take the time to write a nice comment on AF in the near future because you REALLY deserve a big recognition for your work!

PS : I will check a KORG volca FM when possible to compare...

Reply